AI & distribution planning …

It’s been a couple of years now that we hear about Artificial Intelligence (AI), but the last few months it seems we’re shifting to higher gears with #ChatGPT (developed by #OpenAI) made available to the public.

A lot of questions rise about the danger of AI and how we should or shouldn’t use it. This goes as far as official warnings from ex-employees of tech companies that were in charge of developing AI …

I can only assume that the above discussion will be held at governmental level to set the future rules surrounding AI (do’s and don’ts) and will not go further in detail about that.

In the meantime AI could be extremely beneficial to the transport industry, especially to increase planning efficiency when making milk runs (a milk run is a delivery method used to transport mixed loads from various suppliers to one customer. It can also be used to describe a transportation route that has many stops. In this article I’ll point to the last definition).

Current planning methods

The most common way of making a distribution route (read: milk run) is via a semi-automated way, meaning that a TMS (Transport Management System) is used and applies preset parameters.

There are also companies that do manual planning based on printed versions of CMR documents, sequenced on zip-codes and then copy that manual planning into a system when finished — which is even more inefficient — and I would even say a complete waste of resources, time and money.

These ways of planning are far from efficient, let me explain:

• Manual intervention is needed

You’ll need manual intervention as although you can go as deep in detail as you want in terms of setting up the parameters, the system can only do so-much and needs to be checked. The parameters will need to be constantly monitored and, if deemed necessary, adjusted, the final outcome of the planned routes should be equally checked on feasibility.

• The outcome is subjective

Depending on who is doing the planning you’ll get different outcomes. Not one planner will make exactly the same routes. You’re also heavily depending on the knowledge of infrastructure, constraints, assets, customer requirements etc..

• The milk run is static

Once the planning is finished, the milk run will not evolve. It has become a static route that takes the driver from point A to point B and so on and so forth. Almost as-if it’s written in marble.

• External factors are not taken into account

Think about: weather conditions, traffic jams, roadworks, small roads, traffic limitations, etc..

• The same mistake(s) can be made over and over again

Human error is something you’ll have to take into account when using a semi-automated system or a manual planning process. A mistake can be made over and over again if it’s not possible to capture the solution in a system. Changing the planner, lack of knowledge transfer or poorly maintained system parameters can also lead to the sames issues happening more than once.

The planner is the spider in the middle of the web, in which she (or he) will need to procure as much information and details as possible and take that same information into account before making the planning.

For example: lack of knowledge, only partial information available, not being aware of the costs of used assets, any other important factors not transmitted / known etc. will make sure your planning is doom to fail. Your planner is not paying attention to certain constrains, the next day you’ll feel the impact of non-efficient planning in your customer care department …

Conclusion: based on the above, the most used transport planning processes within businesses will not lead to the most efficient or optimized routes. Let alone being cost-efficient.

Although in many cases the difference between making profit or loss is directly linked to the quality of the routes created by the planner. That is a lot of responsibility to put in the hands of planners, especially knowing that the available tools are limited.

But isn’t it all about #data ?

Yes it is. And that’s exactly where AI kicks-in. What if we evolve the role of a transport planner to somebody who is actually feeding the machine ?

Ok I lost you now, bear with me, and let’s roll back a minute.

Remember: in the previously mentioned planning processes, reading in between the lines, it’s clear that a transport planner can only do such-a-good-job as the information / data he possesses. With only half of the information, or complete lack of, it will (again) trigger the transport routes to fail.

So the real question here is, how do we make sure that the transport planner has ALL the information / details she (he) needs, to make a (cost)efficient planning ?

And at the same time, as 3PL, we need to provide accurate, up-to-date information and the right tools (read: systems) to a planner to assure routes that are optimum, consistent and cost efficient.

That is assuming there is still a transport planner making the milk runs.

Let’s take it a step further.

This is the point where Artificial Intelligence can make the difference. AI through machine learning based on historical data (previously created milk runs / distribution routes) and logged issues on those routes can start to learn to predict the most common issues on certain routes and created new routes avoiding these.

Based on the details of each of the assets a 3PL possesses (truck type, available loading meters, maximum loading weight, tail-lift, ADR, etc.), adding different applicable constraints (daily driving period, customer opening times, fixed time slots) the expected result would be a doable and efficient planning.

Add cost information to that (petrol, mileage-fees, human resource, insurance, maintenance costs, write-off etc.) and the result wouldn’t be limited to doable and efficient, each route would be profitable.

Though everything stands or falls with the accuracy of the data. The role of the planner would evolve from actually making the planning to making sure the data available is accurate (and checking the outcome).

If the planner spots an error or non-doable route in the outcome, the planner would need to tell the system the reasons why it’s not doable, so again, through machine learning these reasons are taken into account for the next run.

E.g. the planner would be feeding the machine.

The above is already a major improvement, cost wise as time wise speaking certainly compared to manual and semi-automated planning. The added value of the planner would shift from actually making the planning to assuring data accuracy.

But for the moment the outcome also remains static.

Now only imagine if we would connect real-time traffic information and weather forecasts via the system with the onboard tracking devices, it will become a whole different ballgame.

Dynamic AI route planning …

The one that will be able to combine AI cutting edge tech within a TMS will be miles in front of its competition.

My findings: I strongly believe there is a future for AI within the transport sector. The above is merely a nutshell of the unlimited possibilities and advantages if AI would be used. I’m eager to see how this will evolve and improve the transport sector overall.

Sources and other interesting topic-linked articles
Machine Learning : https://en.wikipedia.org/wiki/Machine_learning
3PL : https://redstagfulfillment.com/3pl-definition-process-resources/
How AI is improving transportation and logistics : https://www.aiplusinfo.com/blog/how-ai-is-improving-transportation-and-logistics/